Speech to text - batch Transcriptins

Question

Speech to text - batch Transcriptins

Mohamed Hussein 710

Good Day,

I've couple of questions regardting speech to text (batch transcritions)

Where's i can find list of supported Models highlighted at the screenshot
I'm getting following error, while locale "ar-EG" is supported as mentioned here https://learn.microsoft.com/en-us/azure/ai-services/speech-service/language-support?tabs=stt

User's image

   {
       "code": "InvalidRequest",
       "message": "The transcription locale does not match the locale of a model.",
       "innerError": {
           "code": "InvalidPayload",
           "message": "The transcription locale does not match the locale of a model."
       }
   }

santoshkc 14,500 Reputation points Microsoft External Staff Moderator

2025-04-25T16:43:13.2866667+00:00

Hi @Mohamed Hussein,

I attempted to reproduce the issue on my end and here are the findings:

I used the same request pattern with locale: "ar-EG" and intentionally included a model that doesn't support Arabic, which returned the same error you received.

Next, I removed the "model" field entirely and allowed the system to auto-select the correct model. The transcription request was accepted successfully:

Please try removing the model field in your transcription request. The service will automatically choose the appropriate model based on the locale (ar-EG in your case), and your request should go through without errors.

I hope this helps. Thank you.
Mohamed Hussein 710 Reputation points

2025-04-25T17:25:44.3766667+00:00
Thank you santoshkc

Appreciate your answer,

Actually, my task is tp compare the accuracy of different available models (include Whisper and Azure stt)

So, I'm looking for a way to state the model

Where shall i list Models IDs, it's only available on the speech playground

How to know who locale supported by each model

Thank you
santoshkc 14,500 Reputation points Microsoft External Staff Moderator

2025-04-29T13:01:14.2066667+00:00

Hi @Mohamed Hussein,

Thank you for the clarification, that you’re comparing the accuracy of different models, including Whisper and Azure STT. In this case, specifying the model explicitly is indeed required. While the full list of model IDs isn’t published in one place, you can view and select available models in your region through the Speech Studio Voice Gallery.

Alternatively, you can retrieve available models programmatically using the GET {endpoint}/speechtotext/v3.2-preview.2/models API, which returns all base and custom models tied to your Azure subscription in the selected region. Unfortunately, the supported locales for each model aren’t listed in the API response.

I hope you understand! Thank you.
santoshkc 14,500 Reputation points Microsoft External Staff Moderator

2025-04-30T14:00:41.63+00:00

Hi @Mohamed Hussein,

Following up to see if the given response was helpful. In case if you have any resolution, please do share that same with the community as it can be helpful to others. Thank you.
Mohamed Hussein 710 Reputation points

2025-04-30T14:13:43.5766667+00:00
Thank you @santoshkc

Appreciate your support

I'm afraid listing APIs returns an empty list,

curl --location 'https://husazaiservices.cognitiveservices.azure.com/speechtotext/v3.2/models' \ --header 'Ocp-Apim-Subscription-Key: ______________'
Manas Mohanty 3,700 Reputation points Microsoft External Staff Moderator

2025-05-05T10:03:52.85+00:00

Hi @Mohamed Hussein,

We are checking to see if the above pointers helped address your issue.

Thank you.
Mohamed Hussein 710 Reputation points

2025-05-05T10:45:31.9+00:00

@Manas Mohanty Thank you, it works. Feel free to close the ticket
Manas Mohanty 3,700 Reputation points Microsoft External Staff Moderator

2025-05-05T14:31:54.09+00:00

Ok @Mohamed Hussein,

Could you accept the below answer if you appreciated the inputs.

Thank you.

Accepted answer

0 additional answers

Your answer

santoshkc 14,500 Reputation points Microsoft External Staff Moderator

2025-04-25T16:43:13.2866667+00:00

Hi @Mohamed Hussein,

I attempted to reproduce the issue on my end and here are the findings:

I used the same request pattern with locale: "ar-EG" and intentionally included a model that doesn't support Arabic, which returned the same error you received.

Next, I removed the "model" field entirely and allowed the system to auto-select the correct model. The transcription request was accepted successfully:

Please try removing the model field in your transcription request. The service will automatically choose the appropriate model based on the locale (ar-EG in your case), and your request should go through without errors.

I hope this helps. Thank you.
Mohamed Hussein 710 Reputation points

2025-04-25T17:25:44.3766667+00:00

Thank you santoshkc

Appreciate your answer,

Actually, my task is tp compare the accuracy of different available models (include Whisper and Azure stt)

So, I'm looking for a way to state the model

Where shall i list Models IDs, it's only available on the speech playground

How to know who locale supported by each model

Thank you
santoshkc 14,500 Reputation points Microsoft External Staff Moderator

2025-04-29T13:01:14.2066667+00:00

Hi @Mohamed Hussein,

Thank you for the clarification, that you’re comparing the accuracy of different models, including Whisper and Azure STT. In this case, specifying the model explicitly is indeed required. While the full list of model IDs isn’t published in one place, you can view and select available models in your region through the Speech Studio Voice Gallery.

Alternatively, you can retrieve available models programmatically using the GET {endpoint}/speechtotext/v3.2-preview.2/models API, which returns all base and custom models tied to your Azure subscription in the selected region. Unfortunately, the supported locales for each model aren’t listed in the API response.

I hope you understand! Thank you.
santoshkc 14,500 Reputation points Microsoft External Staff Moderator

2025-04-30T14:00:41.63+00:00

Hi @Mohamed Hussein,

Following up to see if the given response was helpful. In case if you have any resolution, please do share that same with the community as it can be helpful to others. Thank you.
Mohamed Hussein 710 Reputation points

2025-04-30T14:13:43.5766667+00:00

Thank you @santoshkc

Appreciate your support

I'm afraid listing APIs returns an empty list,

curl --location 'https://husazaiservices.cognitiveservices.azure.com/speechtotext/v3.2/models' \ --header 'Ocp-Apim-Subscription-Key: ______________'
Manas Mohanty 3,700 Reputation points Microsoft External Staff Moderator

2025-05-05T10:03:52.85+00:00

Hi @Mohamed Hussein,

We are checking to see if the above pointers helped address your issue.

Thank you.
Mohamed Hussein 710 Reputation points

2025-05-05T10:45:31.9+00:00

@Manas Mohanty Thank you, it works. Feel free to close the ticket
Manas Mohanty 3,700 Reputation points Microsoft External Staff Moderator

2025-05-05T14:31:54.09+00:00

Ok @Mohamed Hussein,

Could you accept the below answer if you appreciated the inputs.

Thank you.

Answer 1

Hi @Mohamed Hussein,

Sorry for delay in getting working command for listing command.

There was missing keywords in previous command shared from Santosh.

Here are the windows commands

to get list of bases models

curl -X GET "{endpoint}/speechtotext/v3.2-preview.2/models/base" -H "Ocp-Apim-Subscription-Key: REPLACE_WITH_YOUR_SUBSCRIPTION_KEY"

to get list of custom models (will show empty array if no custom model available)

curl -X GET "{endpoint}/speechtotext/v3.2-preview.2/models" -H "Ocp-Apim-Subscription-Key: REPLACE_WITH_YOUR_SUBSCRIPTION_KEY"

Hope it helps.

Thank you

Share via

Speech to text - batch Transcriptins

0 additional answers

Your answer